Forum: help

RE: How to solve Error in maxNRCompute [ Reply ]
By: Arne Henningsen on 2018-03-04 18:08

The log-likelihood function seems to be incorrectly specified or programmed. The error message indicates that the optimisation algorithm checks the parameter vector c(-12.077769, -1.010946, 1.451812). The optimization algorithm was so successful that the log-likelihood function becomes infinity:

ll(c(-12.077769, -1.010946, 1.451812))
[1] Inf

As the log-likelihood function is infinity, the algorithm cannot calculate its gradients / partial derivatives with respect to the parameters and, thus, terminates without success. Hence, you need to correct the specification or the coding of the log-likelihood function.

How to solve Error in maxNRCompute [ Reply ]
By: yap Yap on 2018-01-13 06:27

[forum:45563]

i got error (Error in maxNRCompute(fn = function (theta, fnOrig, gradOrig = NULL, hessOrig = NULL, : NA in gradient) when i want to find the maximum likelihood estimates with the following code. i realized i won't get the error with smaller loop but i will get the error when I use "i=2000".

beta0<--8
beta1<-0.03
gamma<-0.0105
n<-100
beta0hat<-NULL
beta1hat<-NULL
gammahat<-NULL

for (i in 1:2000)
{
u<-runif(n)
u
x<-rnorm(n)
x
c<-rexp(100,1/1515)
c
t1<-(1/gamma)*log(1-((gamma/(exp(beta0+beta1*x)))*(log(1-u))))
t1
t<-pmin(t1,c)
t
delta<-1*(t1>c)
delta
length(delta)
cp<-length(delta[delta==1])/n
cp
delta[delta==1]<-ifelse(rbinom(length(delta[delta==1]),1,0.5),1,2)
delta
#0 uncensored
#1 right censored
#2 interval censored
deltae<-ifelse(delta==0, 1,0)
deltar<-ifelse(delta==1, 1,0)
deltai<-ifelse(delta==2, 1,0)
dat=data.frame(t,delta, deltae,deltar,deltai,x)
dat$interval[delta==2] <- as.character(cut(dat$t[delta==2], breaks=seq(0, 500, 100)))
labs <- cut(dat$t[delta==2], breaks=seq(0, 500, 100))
dat$lower[delta==2]<-as.numeric( sub("\\((.+),.*", "\\1", labs) )
dat$upper[delta==2]<-as.numeric( sub("[^,]*,([^]]*)\\]", "\\1", labs) )
midcut<-function(x,from,to,by){
## cut the data into bins...
x=cut(x,seq(from,to,by),include.lowest=T)
## make a named vector of the midpoints, names=binnames
vec=seq(from+by/2,to-by/2,by)
names(vec)=levels(x)
## use the vector to map the names of the bins to the midpoint values
unname(vec[x])
}
dat$midpoint[delta==2]<-midcut(dat$t[delta==2],0,500,100)
data0<-dat[which(dat$delta==0),]#uncensored data
data1<-dat[which(dat$delta==1),]#right censored data
data2<-dat[which(dat$delta==2),]#interval censored data
library(maxLik)
#without imputataion
ll<-function(para)
{
b0<-para[1]
b1<-para[2]
g<-para[3]
e<-sum((b0+b1*data0$x)+g*data0$t+(1/g)*exp(b0+b1*data0$x)*(1-exp(g*data0$t)))
r<-sum((1/g)*exp(b0+b1*data1$x)*(1-exp(g*data1$t)))
i<-sum((1/g)*exp(b0+b1*data2$x)*(exp(g*data2$upper)-exp(g*data2$lower)))
l<-e+r+i
return(l)
}
est<-maxLik(logLik=ll,start=c(para<-c(-8,0.03,0.0105)))
}